Tuesday, March 28, 2017

Lagged Selection Bias

A good contribution in the battle of fighting cognitive pollution. From The Death of the White Working Class Has Been Greatly Exaggerated by Malcolm Harris. Harris points out that, intentionally or unintentionally, the researches are misleading because they fail to control variables and compare apples and oranges and then exacerbate that issue by disguising the apples-to-oranges issue by using different scales.
Underlying this argument is a series of reports on the immiseration of the white working class and its members’ increasing tendency to die. But while these papers have garnered a lot of attention, there’s good evidence that their conclusions go too far.

The latest version come from the all-star Princeton University economics couple Anne Case and Angus Deaton, writing under the auspices of the Brookings Institution, a centrist think tank. “Mortality and Morbidity in the 21st Century” might sound like a dull title, but the report has inspired breathless headlines, such as: “Why the White Middle Class Is Dying Faster, Explained in 6 Charts” and “Deaths of Despair: The White American Working Class Is Dying Young.” Brookings got in on the hype game themselves with a blog post titled “Working Class White Americans Are Now Dying in Middle Age at Faster Rates Than Minority Groups.” Those headlines are what “Mortality and Morbidity” was designed to elicit; the paper’s leading graphic has served as the peg for most of these stories. Unfortunately, Case and Deaton’s leading graphic is also one of their most misleading.

Harris identifies the primary source of error in the analysis as lagged selection bias. Lagged Selection Bias is when the group you are comparing at the end of a time period differs in some material way from that same group at the beginning.
In the International Journal of Epidemiology, researchers Jennifer B. Dowd and Amar Hamoudi suggested the Olshansky results could reflect increasing high school graduation rates more than increasing mortality. As a greater proportion of Americans finish 12th grade, lagged selection bias (a term that Dowd and Hamoudi coined) means the demographic of non-completers shrinks over time, and the longitudinal comparison gets less valid. “In terms of mortality risk, those excluded from high school in the early part of the 20th century are not comparable with those excluded from high school a generation later,” they write, “because those left behind by the high school expansions in mid-century likely had childhoods that were more disadvantaged along many dimensions, and so were at higher mortality risk all along.”
Exaggerating the numbers makes the lagged selection bias issue clearer. Let's say in 1950, 50% of the white population had a HS diploma or less. In 2017 that percentage has shrunk to 10% of the population because of intense public policies to improve educational outcomes and high school graduation rates. It is reasonable to assume that the remaining 10% have the greatest concentration of handicaps, barriers, and challenges compared to the 40% who are now achieving higher education attainment. Yes, the 50% and the 10% are the same in the sense that they fit the same category of "white HS and less". But they differ in that the HS and less group in 1950 were healthier, smarter, had more intact families, etc. whereas the 10% in 2017 likely came from more impoverished backgrounds, had more fractured family origins, and likely faced higher levels of challenge and discrimination. The 1950 apples are being compared to 2017 oranges.

I think Harris ends with a harsher judgment of the researchers than I would have.
Despite the headlines, when you compare apples to apples, white Americans remain better off on average than black Americans across the board. For example, to fit black and white rates of heart disease mortality on the same graphs, Case and Deaton had to use different scales (see above). Comparing a range of eight deaths per 100,000 in white women to a range of 40 deaths per 100,000 in black women is to pay closer attention to the former. In these graphs, white lives literally count more, and black lives less. But whether in health, income, wealth, or educational attainment, American white privilege is still very much in effect, and no statistical tomfoolery can change that.
Clearly Harris is coming to the table with a particular left-bias with a focus on race. But in this instance, I think both the original researchers and Harris are both correct.

Yes, it is true that white outcomes are still better than black outcomes. It is also possibly true that white outcomes have either plateaued or possibly dipped depending on how you control for lagged selection bias. Harris sees the focus on white outcomes as evidence of "white privilege" but that may be an ideological filter. It is not unreasonable to be more attuned to slight changes in Population X if Population X is five times larger than population Y.

None-the-less his focus on the error in interpretation arising from lagged selection bias is correct and important.

No comments:

Post a Comment